Making the invisible visible: An automated clinical decision support tool for Intimate Partner Violence Risk and Severity Prediction (AIRS)

让不可见变得可见：用于亲密伴侣暴力风险和严重程度预测 (AIRS) 的自动化临床决策支持工具

• 批准号: 10522589
• 负责人: Bharti Khurana
• 金额: $ 84.83万
• 依托单位: BRIGHAM AND WOMEN'S HOSPITAL
• 依托单位国家: 美国
• 财政年份: 2022
• 资助国家: 美国
• 起止时间: 2022-09-21 至 2026-06-30
• 项目状态: 未结题
• 来源: https://reporter.nih.gov/project-details/10522589
• 关键词: Accident and Emergency department; Acute; Age; Algorithms; Automated Clinical Decision Support; Awareness; COVID-19 pandemic; Caring; Chronic; Chronic Disease; Chronology; Clinical; Clinical Data; Clinical/Radiologic; Computerized Medical Record; Cues; Data; Data Set; Decision Support Model; Diagnosis; Early Diagnosis; Early identification; Elderly; Feedback; Feeling; Fracture; Fright; Gender; Goals; Gold; Harvest; Health; High Prevalence; Homicide; Image; Injury; Institution; Intervention; Knowledge; Label; Language; Lead; Life; Location; Machine Learning; Medical; Methods; Modality; Modeling; Neural Network Simulation; Non-accidental; Pathology; Patient Care; Patient Self-Report; Patients; Pattern; Performance; Phase; Physicians; Play; Public Health; Questionnaires; Race; Radiology Specialty; Recording of previous events; Reporting; Research; Resource Sharing; Risk; Risk Factors; Role; Safety; Screening procedure; Sensitivity and Specificity; Severities; Shame; Site; Social Workers; Survivors; System; Testing; Time; Translating; Visit; Visualization; Woman; base; case control; clinical care; clinical decision support; clinical practice; clinical predictors; clinical risk; cohort; craniofacial; evidence base; experience; face bone structure; health care settings; imaging study; improved outcome; interdisciplinary approach; intimate partner violence; limb bone; machine learning algorithm; machine learning model; men; multidisciplinary; multimodality; musculoskeletal injury; novel; pandemic disease; patient population; pediatric trauma; pilot test; point of care; predictive modeling; predictive tools; prevent; psychological violence; radiologist; recurrent neural network; repository; risk stratification; serial imaging; severe injury; sexual violence; social; support tools; telehealth; tertiary care; tool

Project Abstract This project is focused on developing an automated clinical decision support tool for predicting Intimate Partner Violence (IPV) risk and severity based on historical imaging and clinical data. Despite the high prevalence and urgency of this critical public health issue, there is currently no objective tool to diagnose IPV. The challenges in detecting IPV in the health care setting are due to multiple factors, including the patient’s feelings of shame and fear of consequences and physician’s lack of awareness and fear of offending the patient and partner. While imaging plays an essential role in diagnosing nonaccidental trauma in children because of clear well- established patterns of abuse on imaging studies, a lack of evidence-based research on IPV related imaging patterns has led to under-recognition and underdiagnosis of IPV. By recognizing location and imaging patterns specific to IPV on current and previous radiological studies, radiologists can help identify IPV when the victims are not forthcoming. Our hypothesis is that a multidimensional clinical support tool including imaging and clinical findings harvested from the electronic medical record can provide an accurate and comprehensive calculation of IPV risk. The automated IPV risk and severity predictions can then be integrated to transform the care plan for survivors and make the “invisible” visible. Aim 1: To define IPV related imaging patterns and severity by analyzing radiological studies of known IPV survivors and matched controls Aim 2: To determine IPV risk and severity prediction by developing a clinical decision support tool derived from historical imaging and clinical predictors. Aim 3: To validate the IPV prediction model on new datasets and evaluate the integration of results in radiology workflow using a safe repository. If our hypotheses are correct, established IPV related imaging patterns, a CDS tool derived from historical imaging and clinical predictors integrated into clinical care will be able to diagnose IPV objectively.

项目摘要 该项目的重点是开发自动化临床决策支持工具来预测 基于历史影像和临床数据的亲密伴侣暴力 (IPV) 风险和严重程度。 尽管这一重要的公共卫生问题非常普遍且紧迫，但目前还没有 诊断 IPV 的客观工具 在医疗保健环境中检测 IPV 的挑战是 由于多种因素，包括患者的羞耻感和对后果的恐惧， 医生在成像时缺乏意识并担心冒犯患者和伴侣。 由于明确的良好诊断，它在诊断儿童非意外创伤方面发挥着重要作用 影像学研究滥用的既定模式，缺乏 IPV 的循证研究 相关的成像模式导致了 IPV 的识别不足和诊断不足。 当前和之前的放射学研究中 IPV 特有的位置和成像模式， 当受害者没有出现时，放射科医生可以帮助识别 IPV。 多维临床支持工具，包括从实验室收集的影像和临床发现 电子病历可以提供准确、全面的IPV风险计算。 然后可以集成自动化的 IPV 风险和严重程度预测来改变护理方式 为幸存者制定计划，让“看不见的”变得可见。 目标 1：通过分析 IPV 的放射学研究来定义 IPV 相关的成像模式和严重程度 已知的 IPV 幸存者和匹配的对照 目标 2：通过开发临床决策支持来确定 IPV 风险和严重程度预测 来自历史影像和临床预测的工具。 目标 3：在新数据集上验证 IPV 预测模型并评估集成 使用安全存储库实现放射学工作流程。 如果我们的假设是正确的，则建立 IPV 相关成像模式，这是源自于的 CDS 工具 历史影像和临床预测融入临床护理将能够诊断 IPV 客观地。